Learning Coupled Forward-Inverse Models with Combined Prediction Errors
Learning Coupled Forward-Inverse Models with Combined Prediction Errors
复制标题
学习具有组合预测误差的耦合正逆模型
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Jan Peters
中科院分区：
文献类型：
--
作者：
Dorothea Koert;Guilherme J. Maeda;G. Neumann;Jan Peters
非结构化环境中的挑战性任务需要机器人学习复杂的模型。在给定大量信息的情况下，学习多个简单模型可以为单一复杂网络提供一种有效的替代方案。训练多个模型——也就是说，学习它们的参数和它们的职责——已经被证明是非常困难的，因为优化容易出现局部最小值。为了有效地学习不同背景下的多个模型，我们开发了一种基于期望最大化（EM）的新算法。与同类概念相比，该算法同时利用正逆模型和正逆模型的预测误差来训练成对正逆模型的多个模块。特别是，我们表明，我们的方法比只考虑前向模型在逆空间包含多个解的任务上的误差产生了实质性的改进。
Challenging tasks in unstructured environments require robots to learn complex models. Given a large amount of information, learning multiple simple models can offer an efficient alternative to a monolithic complex network. Training multiple models-that is, learning their parameters and their responsibilities-has been shown to be prohibitively hard as optimization is prone to local minima. To efficiently learn multiple models for different contexts, we thus develop a new algorithm based on expectation maximization (EM). In contrast to comparable concepts, this algorithm trains multiple modules of paired forward-inverse models by using the prediction errors of both forward and inverse models simultaneously. In particular, we show that our method yields a substantial improvement over only considering the errors of the forward models on tasks where the inverse space contains multiple solutions.